[capibawa]

What and why

Hello Kitty Party Land 1.0.2 (01001340271EC000, Unity 6000.3.11f1) drew its characters as huge stretched triangles. The shader split reported that the ComputeShader objects (class 72, type hash 43a0948a…) in Data/Resources/unity default resources and Data/globalgamemanagers.assets had no type tree from a bundle and no layout in unityfs.py, so their kernels were left out. The run then logged the dispatched program has no compute shader where Unity's built-in Internal-Skinning dispatch should run, and the characters drew from vertex buffers it never wrote.

COMPUTE_6000_3 is the 2022.3 layout with an int resType after each resource's texDimension and threadGroupSize a staticvector again, written from AssetRipper's 6000.3.11f1 release tree.

Reference entry: a "Unity 6000.3" item under Unity ComputeShader objects, beside the 2017.4, 2019.4 and 2021.3 ones.

Checked

• Tests: dm/gfx/tests/test_unity_known_trees.py (now listing the 6000.3 hash) passes. The GPU tests (make -C dm/gfx/tests GAME=<id> run) were not run.
• Layout: matches all 159 nodes and alignment flags of the 6000.3.11f1 dump. All 11 ComputeShader objects in the two files read to their exact ends (Internal-Skinning 40 kernels, Internal-BlendShape 9).
• Games run: Hello Kitty Party Land 1.0.2 (v131072). The split finds 92 compute shaders instead of 1, the run logs no shader missing from the map and no dispatch without a compute shader, and the characters draw correctly on the Mac.
• Devices and OS: M3 Pro, macOS 27.0.
• Reach to other games: scripts/regress.py check --games was not run. The change only adds a KNOWN_TREES entry for a type hash no layout covered, which the split's second pass uses only for stripped files of that hash. Games of other Unity versions read exactly as before.

Docs

• The finding is in its game note or its docs/reference/<subsystem>.md entry, and python3 dm/tools/docs_symptoms.py was run (the symptom index; an item under an existing entry, so it did not change)
• A chain tool changed: not applicable (unityfs.py reads Unity files for the split; the translation chain is untouched)
• No game files, keys or data extracted from a game in the change or the attachments

[seedds]

Landed on main as ba366d2, with you credited as co-author. Thank you for the clear diagnosis.

The layout is yours. It is the same expression as in #791, so the commit consolidates the two: it adds #791's decoding fixture, and the docs entry keeps your Hello Kitty finding.

I compared the layout with AssetRipper's 6000.3.0b4, 6000.3.0f1, 6000.3.11f1 and 6000.3.20f1 dumps, and all 159 nodes and alignment flags match. I don't have a game with this hash here, so the docs record your game results as reported by this PR, not repeated.